This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Are individual-mouse-based statistical tests possible with CellChat?

Hey people,

I asked this question on a few other forums too, but I think my chances of getting an answer on any of single forum are modest, so I want to ask here too, if that's okay.

Anyway, to the point: using CellChat (v2.2.0), when comparing between two groups (WT vs. KO,), each of which comprised of 4-5 mice, is there a way to find out if a specific pathway is enriched in one group (compared to the other) in a specific source cell type and a specific target cell type in a statistically-significant manner, based on the individual mice (comparing e.g. 4 values for "WT" [4 mice] and 4 values for "KO", thus being able to run a statistical test)?

I'll try to explain it by example, using, in this case, the "rankNet" function - but I am totally fine with any other function if it can help.When I run:

gg1 <- rankNet(cellchat, 
               mode = "comparison", 
               signaling = "COLLAGEN",
               sources.use = "FIB1",
               targets.use = "FIB2",
               do.stat = TRUE,
               return.data = TRUE)

The p-values (in "gg1[["signaling.contribution"]]$pvalues") will always be "0" (or completely absent), no matter which source, target and pathway are specified. But that, to my understanding, is because the source code for "rankNet" forces a "pvalue" of "0" here, because there are only 2 "prob.values" (one for KO, one for WT). However, as I've mentioned, my WT and KO groups consist of 4-5 mice each. Is there a way to leverage that fact to be able to find out if a specified pathway is substantially changed between the WT and KO groups in a specified source and a specified target?

communication interaction cellchat cell-cell ligand-receptor interactome
• 481 views • link
updated by ATpoint ATpoint • written by omer.shomrat • omer.shomrat

You would need to run CellChat on every WT-KO individual mice pair to find that. I wonder whether it is meaningful to do that since usually you would want significant results that hold true in your group comparison, not in individual mice but not the group. What I would do (and have done before) is to get the LR scores per individual mouse (for example running CellChat or other frameworks individually per mouse) and then do the actual differential analysis outside of the tool. For example, I compared the LR scores with limma, or any other statistical test, and be it a t-test if you have enough mice to get results.

• 0 views • link
written by ATpoint ATpoint

0 answers

No answers yet.

Log in to answer this question.

More posts like this